1 Direct Access for files
2 -----------------------
7 The page cache is usually used to buffer reads and writes to files.
8 It is also used to provide the pages which are mapped into userspace
11 For block devices that are memory-like, the page cache pages would be
12 unnecessary copies of the original storage. The DAX code removes the
13 extra copy by performing reads and writes directly to the storage device.
14 For file mappings, the storage device is mapped directly into userspace.
20 If you have a block device which supports DAX, you can make a filesystem
21 on it as usual. The DAX code currently only supports files with a block
22 size equal to your kernel's PAGE_SIZE, so you may need to specify a block
23 size when creating the filesystem.
25 Currently 3 filesystems support DAX: ext2, ext4 and xfs. Enabling DAX on them
29 -----------------------------
31 When mounting the filesystem, use the "-o dax" option on the command line or
32 add 'dax' to the options in /etc/fstab. This works to enable DAX on all files
33 within the filesystem. It is equivalent to the '-o dax=always' behavior below.
36 Enabling DAX on xfs and ext4
37 ----------------------------
42 1. There exists an in-kernel file access mode flag S_DAX that corresponds to
43 the statx flag STATX_ATTR_DAX. See the manpage for statx(2) for details
44 about this access mode.
46 2. There exists a persistent flag FS_XFLAG_DAX that can be applied to regular
47 files and directories. This advisory flag can be set or cleared at any
48 time, but doing so does not immediately affect the S_DAX state.
50 3. If the persistent FS_XFLAG_DAX flag is set on a directory, this flag will
51 be inherited by all regular files and subdirectories that are subsequently
52 created in this directory. Files and subdirectories that exist at the time
53 this flag is set or cleared on the parent directory are not modified by
54 this modification of the parent directory.
56 4. There exist dax mount options which can override FS_XFLAG_DAX in the
57 setting of the S_DAX flag. Given underlying storage which supports DAX the
60 "-o dax=inode" means "follow FS_XFLAG_DAX" and is the default.
62 "-o dax=never" means "never set S_DAX, ignore FS_XFLAG_DAX."
64 "-o dax=always" means "always set S_DAX ignore FS_XFLAG_DAX."
66 "-o dax" is a legacy option which is an alias for "dax=always".
67 This may be removed in the future so "-o dax=always" is
68 the preferred method for specifying this behavior.
70 NOTE: Modifications to and the inheritance behavior of FS_XFLAG_DAX remain
71 the same even when the filesystem is mounted with a dax option. However,
72 in-core inode state (S_DAX) will be overridden until the filesystem is
73 remounted with dax=inode and the inode is evicted from kernel memory.
75 5. The S_DAX policy can be changed via:
77 a) Setting the parent directory FS_XFLAG_DAX as needed before files are
80 b) Setting the appropriate dax="foo" mount option
82 c) Changing the FS_XFLAG_DAX flag on existing regular files and
83 directories. This has runtime constraints and limitations that are
84 described in 6) below.
86 6. When changing the S_DAX policy via toggling the persistent FS_XFLAG_DAX
87 flag, the change to existing regular files won't take effect until the
88 files are closed by all processes.
94 There are 2 per-file dax flags. One is a persistent inode setting (FS_XFLAG_DAX)
95 and the other is a volatile flag indicating the active state of the feature
98 FS_XFLAG_DAX is preserved within the filesystem. This persistent config
99 setting can be set, cleared and/or queried using the FS_IOC_FS[GS]ETXATTR ioctl
100 (see ioctl_xfs_fsgetxattr(2)) or an utility such as 'xfs_io'.
102 New files and directories automatically inherit FS_XFLAG_DAX from
103 their parent directory _when_ _created_. Therefore, setting FS_XFLAG_DAX at
104 directory creation time can be used to set a default behavior for an entire
107 To clarify inheritance, here are 3 examples:
112 xfs_io -c 'chattr +x' a
122 xfs_io -c 'chattr +x' a
131 xfs_io -c 'chattr +x' c
138 The current enabled state (S_DAX) is set when a file inode is instantiated in
139 memory by the kernel. It is set based on the underlying media support, the
140 value of FS_XFLAG_DAX and the filesystem's dax mount option.
142 statx can be used to query S_DAX. NOTE that only regular files will ever have
143 S_DAX set and therefore statx will never indicate that S_DAX is set on
146 Setting the FS_XFLAG_DAX flag (specifically or through inheritance) occurs even
147 if the underlying media does not support dax and/or the filesystem is
148 overridden with a mount option.
152 Implementation Tips for Block Driver Writers
153 --------------------------------------------
155 To support DAX in your block driver, implement the 'direct_access'
156 block device operation. It is used to translate the sector number
157 (expressed in units of 512-byte sectors) to a page frame number (pfn)
158 that identifies the physical page for the memory. It also returns a
159 kernel virtual address that can be used to access the memory.
161 The direct_access method takes a 'size' parameter that indicates the
162 number of bytes being requested. The function should return the number
163 of bytes that can be contiguously accessed at that offset. It may also
164 return a negative errno if an error occurs.
166 In order to support this method, the storage must be byte-accessible by
167 the CPU at all times. If your device uses paging techniques to expose
168 a large amount of memory through a smaller window, then you cannot
169 implement direct_access. Equally, if your device can occasionally
170 stall the CPU for an extended period, you should also not attempt to
171 implement direct_access.
173 These block devices may be used for inspiration:
174 - brd: RAM backed block device driver
175 - dcssblk: s390 dcss block device driver
176 - pmem: NVDIMM persistent memory driver
179 Implementation Tips for Filesystem Writers
180 ------------------------------------------
182 Filesystem support consists of
183 - adding support to mark inodes as being DAX by setting the S_DAX flag in
185 - implementing ->read_iter and ->write_iter operations which use dax_iomap_rw()
186 when inode has S_DAX flag set
187 - implementing an mmap file operation for DAX files which sets the
188 VM_MIXEDMAP and VM_HUGEPAGE flags on the VMA, and setting the vm_ops to
189 include handlers for fault, pmd_fault, page_mkwrite, pfn_mkwrite. These
190 handlers should probably call dax_iomap_fault() passing the appropriate
191 fault size and iomap operations.
192 - calling iomap_zero_range() passing appropriate iomap operations instead of
193 block_truncate_page() for DAX files
194 - ensuring that there is sufficient locking between reads, writes,
195 truncates and page faults
197 The iomap handlers for allocating blocks must make sure that allocated blocks
198 are zeroed out and converted to written extents before being returned to avoid
199 exposure of uninitialized data through mmap.
201 These filesystems may be used for inspiration:
202 - ext2: see Documentation/filesystems/ext2.rst
203 - ext4: see Documentation/filesystems/ext4/
204 - xfs: see Documentation/admin-guide/xfs.rst
207 Handling Media Errors
208 ---------------------
210 The libnvdimm subsystem stores a record of known media error locations for
211 each pmem block device (in gendisk->badblocks). If we fault at such location,
212 or one with a latent error not yet discovered, the application can expect
213 to receive a SIGBUS. Libnvdimm also allows clearing of these errors by simply
214 writing the affected sectors (through the pmem driver, and if the underlying
215 NVDIMM supports the clear_poison DSM defined by ACPI).
217 Since DAX IO normally doesn't go through the driver/bio path, applications or
218 sysadmins have an option to restore the lost data from a prior backup/inbuilt
219 redundancy in the following ways:
221 1. Delete the affected file, and restore from a backup (sysadmin route):
222 This will free the filesystem blocks that were being used by the file,
223 and the next time they're allocated, they will be zeroed first, which
224 happens through the driver, and will clear bad sectors.
226 2. Truncate or hole-punch the part of the file that has a bad-block (at least
227 an entire aligned sector has to be hole-punched, but not necessarily an
228 entire filesystem block).
230 These are the two basic paths that allow DAX filesystems to continue operating
231 in the presence of media errors. More robust error recovery mechanisms can be
232 built on top of this in the future, for example, involving redundancy/mirroring
233 provided at the block layer through DM, or additionally, at the filesystem
234 level. These would have to rely on the above two tenets, that error clearing
235 can happen either by sending an IO through the driver, or zeroing (also through
242 Even if the kernel or its modules are stored on a filesystem that supports
243 DAX on a block device that supports DAX, they will still be copied into RAM.
245 The DAX code does not work correctly on architectures which have virtually
246 mapped caches such as ARM, MIPS and SPARC.
248 Calling get_user_pages() on a range of user memory that has been mmaped
249 from a DAX file will fail when there are no 'struct page' to describe
250 those pages. This problem has been addressed in some device drivers
251 by adding optional struct page support for pages under the control of
252 the driver (see CONFIG_NVDIMM_PFN in drivers/nvdimm for an example of
253 how to do this). In the non struct page cases O_DIRECT reads/writes to
254 those memory ranges from a non-DAX file will fail (note that O_DIRECT
255 reads/writes _of a DAX file_ do work, it is the memory that is being
256 accessed that is key here). Other things that will not work in the
257 non struct page case include RDMA, sendfile() and splice().